Converting XY to ORTF during post?

[Tom McC]

Still find myself mulling over all this stuff.

In principle one could map an XY to an ORTF if the following restrictions were adhered to:
- mics of identical or accurately known gain, and ideal or accurately known cardioid patterns
- hall with very little wall-, floor- or roof reflections (yuk, anechoic city)
- sounds come preferably from the front arc of the mics included angle

But it would need a delay unit or plugin whose delay is dynamically driven by a signal amplitude. Does such a beast exist?

For illustration, with 120° XY:
- L-mic delivers a signal of, say, X dB. At that same instant, the R-mic signal is, say, 4.47dB lower
- the unique way for that condition (L - R = 4.47dB) to be satisfied is for the sound source to have been at 25° to left of front-centre - because it would then be 35° off-axis to the cardioid L-mic (thus "0.82dB down"), while 85° off-axis to the R-mic (5.29dB down)
- with the sound source direction now established, one can then do the geometry to calculate the arrival time differences for a 17 cm spaced array at the same spot. And the rest is history. :-)

So by means of some mapping algorithm, the instantaneous channel difference signal is harnessed to dictate the delay to be applied, probably in post.

Ach, with all the restrictions and equipment tolerances involved, I'm betting it would sound mediocre.

[leddy]

It all sounds like stereo alchemy.

[Richard Black]

A Blumlein shuffler will do it, or something close to it.

[Tom McC]

Originally Posted by Richard Black

A Blumlein shuffler will do it, or something close to it.

A Blumlein shuffler is commonly used to give a bass boost - perhaps 2-3dB between 250 650 Hz - to the S channel, in order to widen the too-narrow imaging of bass sound sources and make them snap into register with their higher frequency brethren. As such, it can indeed improve the imaging of coincident XY recordings (I've had mixed results applying it atop a bare ORTF recording)

But the OP did not actually ask how in general to improve an XY recording. His specific question was how to make XY behave like an ORTF...which was what I was trying to address.

[Petrus]

Originally Posted by Tom McC

In principle one could map an XY to an ORTF if the following restrictions were adhered to:
- mics of identical or accurately known gain, and ideal or accurately known cardioid patterns
- hall with very little wall-, floor- or roof reflections (yuk, anechoic city)
- sounds come preferably from the front arc of the mics included angle

But it would need a delay unit or plugin whose delay is dynamically driven by a signal amplitude. Does such a beast exist?

For illustration, with 120° XY:
- L-mic delivers a signal of, say, X dB. At that same instant, the R-mic signal is, say, 4.47dB lower
- the unique way for that condition (L - R = 4.47dB) to be satisfied is for the sound source to have been at 25° to left of front-centre - because it would then be 35° off-axis to the cardioid L-mic (thus "0.82dB down"), while 85° off-axis to the R-mic (5.29dB down)
- with the sound source direction now established, one can then do the geometry to calculate the arrival time differences for a 17 cm spaced array at the same spot. And the rest is history. :-)

You make one fatal simplification: sound source distance in your sample is geometrically at infinity. We can not know from the intensity difference calculation how far the sound source is, if you care to make some sketches you will notice that the amplitude difference angle would actually discribe an arch with a bend near the microphones and streighthen out quite fast with growing distance. We have no way of knowing which combination of distance and angle is the correct one for the each identical signal pair.

I'll quote my own post few days ago: "There could be a way of measuring the intesities of similar signals and applying different delays depending on the intesity differencies (which should be related to the angular differencies, at least in a recordings done in an anechoic chamber). This requires a bit more than just delaying one track, if it is possible at all. Complicated DSP in any case and the end result woudl likely to be worse than the original." Thus I have the patent rights for this invention, which would produce unusable muddled up mess out of a clear XY recordings.

[Tom McC]

Originally Posted by Petrus

You make one fatal simplification: sound source distance in your sample is geometrically at infinity. We can not know from the intensity difference calculation how far the sound source is, if you care to make some sketches you will notice that the amplitude difference angle would actually discribe an arch with a bend near the microphones and streighthen out quite fast with growing distance. We have no way of knowing which combination of distance and angle is the correct one for the each identical signal pair.

Petrus, it was your perceptive earlier posts that had got me musing on all this. Thanks.

And yes, there was a flaw in my reasoning - I had a gnawing feeling that it was too good to be true :-) But I'm unsure it's as how you described it. For I did not fully follow your "...amplitude difference angle would actually discribe an arch with a bend near the microphones and streighthen out quite fast with growing distance." Maybe you could elaborate on that.

In any case, it won't hurt if I 'fess up my take on the fatal flaw (as perhaps we are saying the same thing):

The anechoic polar response of a mic is surely independent of the emitted intensity of the sound source? (Otherwise the DPA technician measuring the pattern of, say, a 4011 mic would get a different graph if he turned down the strength of his signal generator by 6dB, tantamount to moving the generator further away.) And if one can measure accurately enough, there is, for a cardioid, a unique solution - with no "aliasing" - to the question: "supply a pair of off-axis angles on the same side of the polar plot that differ by a precise dB value of...".

The direction of the sound source can thus unambigously nailed - e.g. the 35° right of the L-mic and 85° left of the R-mic, in my example. The flaw is in the subsequent mapping of this directional info back into arrival time differences at a 17cm spaced ORTF pair. For we could be dealing with a _point source_ of sound, rather than a broad source of parallel waves. And in the case of a point source, the relative path length difference varies depending how far the source is moved back along the established direction line.

Conventional distance cues are, of course, direct to reverberant ratio etc. etc. - but hey, we're discussing yukky semi-anechoic recordings here! With already so many restrictions on hall conditions and equipment tolerances, this virtual ORTF child was already pretty much stillborn without any further fatal simplifications on my part. :-)

[jimjazzdad]

Wow, I wish I had paid more attention in physics class 40 years ago! Fascinating discussion. Of course it begs the question: "if you want ORTF, why not use ORTF? And if Blumlein XY is too clinical, why not MS? Of course, if you can afford a Soundfield or a Tetra..." In my limited reading and understanding of MG, the whole raison d'être for ambisonics was to provide the flexibility to overcome the the limitations of coincident pairs of mics. If I understand it correctly, you can do everything you want with just phase and amplitude, once you eliminate that pesky time difference thing. I really think the world of sound is divided into coincident and spaced capture. ORTF, NOS, and the like were just invented to sell caridiod pencil mics =) (just kidding, right?)

[apple-q]

Originally Posted by jimjazzdad

In my limited reading and understanding of MG, the whole raison d'être for ambisonics was to provide the flexibility to overcome the the limitations of coincident pairs of mics.

Soundfield is coincident. All the capsules are in the same place so it´s not based on Δ t

It´s about as coicident as MS, XY, Blümlein.

The limitations of coicident stereo can only be overcome by using spaced mics which is why ORTF and NOS were introduced to make a compromise between spaced and coicident setups.

[Petrus]

Originally Posted by Tom McC

Petrus, it was your perceptive earlier posts that had got me musing on all this. Thanks.

And yes, there was a flaw in my reasoning - I had a gnawing feeling that it was too good to be true :-) But I'm unsure it's as how you described it. For I did not fully follow your "...amplitude difference angle would actually discribe an arch with a bend near the microphones and streighthen out quite fast with growing distance." Maybe you could elaborate on that.

If my high school math still holds a fixed artificial delay calculated from the angle calculated from the amplitude difference would discribe a hyperbola as the possible location for the sound source, not a straight line with fixed angle. Thus in a real ORTF the phase difference (delay) between the signals depends also on the sound source distance, not only the angle. This information can not be synthesized from the amplitude difference from XY mic pair.

[RobAnderson]

Whoops - looks like I posted at the same time as everyone else. I'll leave the post, but I came late to the party

MS or Ambisonics cannot really overcome the "limitation" of coincident techniques, namely that there is no time difference between the channels.

It is this small time (read: phase) difference that really makes near-coincident recordings pop, and it can't be faked by simply adding more "out-of-phase" side channel to the recording.

Tom McC's proposal is intriguing, but I think petrus is correct here - the math is a little over my head, but while we could theoretically calculate the source angle using the intensity differences present in each channel, changing the source distance on that vector from the center of the microphone array would change its angle of incidence relative to each (theoretical) microphone in the array, which I believe would result in differing time cues for each possible source distance on that angle of incidence.

[apple-q]

Originally Posted by Petrus

If my high school math still holds a fixed artificial delay calculated from the angle calculated from the amplitude difference would discribe a hyperbola as the possible location for the sound source, not a straight line with fixed angle. Thus in a real ORTF the phase difference (delay) between the signals depends also on the sound source distance, not only the angle. This information can not be synthesized from the amplitude difference from XY mic pair.

Well said. The delay is different for EVERY single sound source and every position.

[Richard Black]

Sorry, should think before typing - a Blumlein shuffler turns ORTF into XY, not the other way around. DOH!

[Tom McC]

Originally Posted by Petrus

If my high school math still holds a fixed artificial delay calculated from the angle calculated from the amplitude difference would discribe a hyperbola as the possible location for the sound source, not a straight line with fixed angle......

Thanks for clarifying, Petrus. It all falls into place now.

A refresher dip in Wikipedia clinched it, with the definition of a hyperbola as expressed in terms of distance from foci.

"A hyperbola may be defined equivalently as the locus of points where the difference of the distances to the two foci is a constant equal to 2a, the distance between its two vertices."

We could rephrase that as "a hyperbola is the curve representing all allowable sound-source locations that would still result in the same constant time delay (path length difference) being experienced between the ORTF capsules (foci)"

[Tom McC]

Originally Posted by Richard Black

Sorry, should think before typing - a Blumlein shuffler turns ORTF into XY, not the other way around. DOH!

Shuffling is a complex subject with more to it than the 'spatial equalization' (Griesinger's term) that I referred to, for tightening up the imaging of XY recordings .To my mind Michael Gerzon's papers and articles give the best treatment.of shuffling. Erudite, thorough, highly recommended.

As I understood it, the principle of shuffling could have been born partly out of necessity: Blumlein was obliged to create a virtual Fig-8 mic by applying his shuffling circuit to two closely-spaced omni capsules! (We easily forget that commercial, high-quality Fig-8 mics were simply unavailable at the time he did his research.)

[Ry-Fi]

Originally Posted by joelpatterson

But... maybe if you delayed both tracks....

Originally Posted by joelpatterson

Of course... the obvious disadvantage with delaying the whole thing is that you'd have to sit around, waiting, to listen to it....

Joel, you crack me up! Some people may not get your humour, but I love it!

[RobAnderson]

Originally Posted by Tom McC

As I understood it, the principle of shuffling could have been born partly out of necessity: Blumlein was obliged to create a virtual Fig-8 mic by applying his shuffling circuit to two closely-spaced omni capsules! (We easily forget that commercial, high-quality Fig-8 mics were simply unavailable at the time he did his research.)

Not sure that this was the case. In his patent, Blumlein clearly describes both pressure (omni) and velocity (ribbon mics), and defines velocity mics as:

"Velocity or moving conductor microphones (e.g. moving strip)...These microphones give a response varying as the cosine of the angle of incidence..."

As you no doubt know, cosine of the angle of incidence describes a figure-of-eight pattern.

And in describing the Mid-Side Technique:

"(53) (1) Two velocity microphones are placed one with its axis of maximum response directly facing in the direction of the centre of the scene, and the other with its axis at right angles to that direction. Both moving strips are in line, and arranged so that this line is vertical, whereas the sound source moves in a horizontal plane. A performer speaking from the middle of the scene will affect only the face-on microphone, but if he moves to one side both microphones will provide outputs, while if he moves the other way similar outputs are provided but the phase of the edge-on microphone is reversed. Since the microphones are close together no phase differences are experienced between them and if their outputs are summed and differenced after a suitable amount of relative amplification the two final channels differ in magnitude in the correct manner for operating the loudspeakers to give the desired directional effect."

Both of these quotes are from Streicher, Ron & Everest, F. Alton. The New Stereo Soundbook. Audio Engineering Assoc. 1998. p. A12

[joelpatterson]

Originally Posted by Ry-Fi

Joel, you crack me up! Some people may not get your humour, but I love it!

[jimjazzdad]

Originally Posted by apple-q

Soundfield is coincident. All the capsules are in the same place so it´s not based on Δ t

It´s about as coicident as MS, XY, Blümlein.

The limitations of coicident stereo can only be overcome by using spaced mics which is why ORTF and NOS were introduced to make a compromise between spaced and coicident setups.

Yes, Soundfield IS coincident, but the limitations MG overcame when he came up with ambisonics were the 2-dimensional characteristics of Blumlein PAIRS. The fact that there is no (well, almost no) time difference between the capsules in a tetrahedral array means the flexibilty to adjust the phase and relative gains to give just about any shape of pick up pattern imaginable. My point is that as soon as the delta-T is introduced, the options of manipulating the soundfield (small s) post recording are much more complicated and limited. I do find the noodling in this thread fascinating tho...but, if you want to mess with the soundfield, go coincident IMO.

That said, I am a big fan of ORTF recordings - sometimes the sound and the maths are in a different relationship than the obvious. to paraphrase Joel, "you have hardly any wait for the sound".

[Tom McC]

Originally Posted by RobAnderson

Not sure that this was the case. In his patent, Blumlein clearly describes both pressure (omni) and velocity (ribbon mics), and defines velocity mics as:........

Rob, one credible source of my statement that Blumlein even commandeerd his shuffling to derive a virtual Fig-8 was the well-regarded J.Audio Eng. Soc. paper of Stanley P. Lipshitz :
"STEREO MICROPHONE TECHNIQUES...Are the Purists Wrong?" (1.)

Here is a relevant section of text lifted from pp 720-1 (hope the moderator is cool with that :-))

"...Blumlein's scheme relied on the realization that simple level differences at the loudspeakers would create both level and phase differences at the listener's ears because of the fact that each ear hears both loudspeakers and the ears are laterally separated (4). By suitably choosing these level differences it is possible to produce stable images between the loudspeakers. This procedure is, for obvious reasons, called "intensity stereo." We would nowadays produce the required level encoding by the use of a coincident pair of directional microphones, classically the "Blumlein pair" of 90°-angled figure-of-eight microphones. Blumlein indeed patented both this arrangement as well as the sum-difference or M-S (mid-side schemes for deriving these loudspeaker feed signals (6), (7) once directional microphones becanme available - specifically the RCA ribbon microphone of the early 1930's. In his initial experiments (4), however, in the absence of other than omnidirectional microphones, he was forced to derive a directional pickup pattern by subtracting and re-equalizing the outputs of two closely spaced pressure microphones, a process he called "shuffling."

So we're not merely dealing with a "factoid" or a "Gearslutz fact" here. :-) But if you nontheless prefer to see the facts first hand, "from the horses mouth", then I urge you to revisit the Blumlein Patents.(no. 4). I did...and lived to tell the tale. :-) And apologies to the OP for further derailing this thread...but I think that had run its course anyway.

=======
ref 1 J. Audio Eng Soc, Vol 34, pp 716-744 (1986)
sub-ref 4 A.D. Blumlein. Brit.Pat. 394325 (appl. Dec 14,1931);
[reprint J.Audio Eng. Soc.Vol 6, pp 91-98, 130 (1958)]
sub-ref 6 A.D. Blumlein. Brit.Pat. 429054 (appl. Feb 10, 1934)
sub-ref 7 A.D. Blumlein. Brit.Pat. 456444 (appl. Feb 7, 1935)
=======

[RobAnderson]

Sorry Tom!

I misunderstood what you were saying in your post. I (incorrectly) thought you were referring to shuffling as it pertains to M-S in the 1931 patent.

My bad...

But the quote in my earlier post is, in fact, directly from the 1931 patent, and since he (Blumlein) is describing velocity and not pressure mic's there, I would assume that he had available ribbon mic's by that time.